Skip to content

MTP Model failed: failed to load MTP head from 'Qwen3.6-27B-Q4_K_M-mtp.gguf' #6

Description

@MyBirdVery6

Name and Version

I used the following two models:
https://huggingface.co/froggeric/Qwen3.6-27B-MTP-GGUF/tree/main
https://huggingface.co/unsloth/Qwen3.6-27B-MTP-GGUF/tree/main
However, an error occurred during startup. The error log is as follows:

print_info: freq_base_train = 10000000.0
print_info: freq_scale_train = 1
print_info: n_ctx_orig_yarn = 262144
print_info: rope_yarn_log_mul = 0.0000
print_info: rope_finetuned = unknown
print_info: mrope sections = [11, 11, 10, 0]
print_info: model type = ?B
print_info: model params = 27.32 B
print_info: general.name = Qwen3.6 27B
print_info: vocab type = BPE
print_info: n_vocab = 248320
print_info: n_merges = 247587
print_info: BOS token = 248044 '<|endoftext|>'
print_info: EOS token = 248046 '<|im_end|>'
print_info: EOT token = 248046 '<|im_end|>'
print_info: PAD token = 248044 '<|endoftext|>'
print_info: LF token = 198 'Ċ'
print_info: FIM PRE token = 248060 '<|fim_prefix|>'
print_info: FIM SUF token = 248062 '<|fim_suffix|>'
print_info: FIM MID token = 248061 '<|fim_middle|>'
print_info: FIM PAD token = 248063 '<|fim_pad|>'
print_info: FIM REP token = 248064 '<|repo_name|>'
print_info: FIM SEP token = 248065 '<|file_sep|>'
print_info: EOG token = 248044 '<|endoftext|>'
print_info: EOG token = 248046 '<|im_end|>'
print_info: EOG token = 248063 '<|fim_pad|>'
print_info: EOG token = 248064 '<|repo_name|>'
print_info: EOG token = 248065 '<|file_sep|>'
print_info: max token length = 256
load_tensors: loading model tensors, this can take a while... (mmap = false, direct_io = false)
llama_model_load: error loading model: invalid vector subscript
�[0mllama_model_load_from_file_impl: failed to load model
�[0msrv load_model: failed to load MTP head from 'Qwen3.6-27B-Q4_K_M-mtp.gguf'
�[0msrv operator (): operator (): cleaning up before exit...
main: exiting due to model loading error
�[0mPress any key to continue . . .

The graphics card I'm using is an RTX 4080 Super with 16GB of VRAM with CUDA 12.4, and I have 32GB of system RAM.Below are my startup configuration parameters:

@echo off
echo.

cd /d "D:\mtp-turboquant"

set CUDA_SCALE_LAUNCH_QUEUES=4x

.\llama-server.exe ^
-m Qwen3.6-27B-Q4_K_M-mtp.gguf ^
--mmproj mmproj-Qwen3.6-27B-f16.gguf ^
-fa on ^
--spec-type mtp --spec-draft-n-max 2 ^
-ctk q8_0 -ctv turbo3 ^
--n-gpu-layers 99 ^
--temp 0.1 --top-k 20 ^
--flash-attn on ^
--reasoning off ^
--no-mmap --mlock ^
--reasoning-format deepseek ^
--repeat-penalty 1.15 ^
--presence-penalty 1.5 ^
--frequency-penalty 0.0 ^
-np 1 ^
-t 8 ^
--host 0.0.0.0 --port 8189 ^
-c 156888 -b 2048 -ub 512 ^
--jinja ^
--chat-template-file "3.6_chat_template-v10.jinja"

pause

My compilation script is:

@echo off
cd /d "D:\Cmake\llamacpp-MTP-TurboQuant"

cmake -B build ^
-G "Visual Studio 17 2022" -A x64 -T "cuda=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.4" ^
-DGGML_CUDA=ON ^
-DCMAKE_CUDA_ARCHITECTURES="89" ^
-DGGML_CUDA_FA_ALL_QUANTS=ON ^
-DGGML_NATIVE=OFF ^
-DCMAKE_BUILD_TYPE=Release

cmake --build build --config Release -j 14 --clean-first --target llama-cli llama-mtmd-cli llama-server llama-gguf-split

So, please help me see where I went wrong?

Operating systems

Windows

GGML backends

CUDA

Hardware

14700KF+RTX4080Super+32GB RAM
D:\mtp-turboquant>nvidia-smi
Sun May 24 22:07:27 2026
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 551.78 Driver Version: 551.78 CUDA Version: 12.4 |
|-----------------------------------------+------------------------+----------------------+
| GPU Name TCC/WDDM | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA GeForce RTX 4080 ... WDDM | 00000000:01:00.0 Off | N/A |
| 30% 29C P8 14W / 320W | 582MiB / 16376MiB | 8% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| 0 N/A N/A 5664 C+G ...5n1h2txyewy\ShellExperienceHost.exe N/A |
| 0 N/A N/A 6784 C+G C:\Windows\explorer.exe N/A |
| 0 N/A N/A 7376 C+G ....Search_cw5n1h2txyewy\SearchApp.exe N/A |
| 0 N/A N/A 8536 C+G ...CBS_cw5n1h2txyewy\TextInputHost.exe N/A |
| 0 N/A N/A 8904 C+G ...ky-borders-v1.3.1\tacky-borders.exe N/A |
| 0 N/A N/A 10620 C+G D:\v2rayN-windows-64\v2rayN.exe N/A |
| 0 N/A N/A 10704 C+G D:\Chrome++\App\chrome.exe N/A |
| 0 N/A N/A 11188 C+G ...ms@opencode-aidesktop\OpenCode.exe N/A |
+-----------------------------------------------------------------------------------------+

Models

Qwen3.6-27B-Q4_K_M-mtp.gguf

Problem description & steps to reproduce

failed to load MTP head from 'Qwen3.6-27B-Q4_K_M-mtp.gguf'

First Bad Commit

No response

Relevant log output

Logs

Metadata

Metadata

Assignees

No one assigned

    Labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions